ietf-smtp
[Top] [All Lists]

Re: BATV breaks rfc2821bis?

2008-05-20 19:53:24


As Ned said, this is a major deviation from what an envelope
sender address used to be for some decades, and it's IMO not
*obvious* how it works in conjunction with RFC 3834 + SIEVE.

How far back do you want to go?  People have been using subaddresses
like user+blah(_at_)domain since 1991.

Try the mid 1980s instead. Our implementation of subaddresses dates back 1990
or so, but we stole the idea from the Andrew Messaging System, which started
out in 1983. I don't know at what point along the way subaddress support was
added, but it was early enough for us to copy it.

Mailing lists have been using VERP
schemes to encode per-message and per-recipient information into the
bounce address at least since 1998.

But these are semantically very different usages from what's being proposed
here. In  parficular, as Carl pointed out, VERP is used on outgoing list
messages, and bouncing them when they hit another list is something of a
feature.

Subaddresses are a little closer, but even here the primary use is in list
subsccription addresses. Perhaps some people use them to implement a sort of
BATV-like scheme, but I have a hard time believing it is a common practice.

The advice in 3834 to send vacation responses to the bounce address
was already dubious in 2004, since by then there was lots of list
management software that interpreted anything sent back to a VERP
address as a bounce.  (Yes, I know that a sensible vacation program
that followed all the advice in 3834 wouldn't respond to list mail at
all, but as we all know, life is not that tidy.)

Well, all I can say is that back in 2000 or so we inherited some code that sent
these to the header From: address. That was a customer complaint generator.
Once we switched it to use the envelope from, these complaints basically just
stopped. So I guess my question is: If not the envelope from address, then
what?

FWIW, the ongoing problem we have with vacation messages is the "don't send
unless the recipient's address actually appears in the header" rule. This
doesn't  interact well with autoforwarders. We provide a means to amend the
list of addresses a user "owns", but it seems it's too much to ask people to
set this up.

                                Ned